The Japanese ASJ RTN-Model 2018, European Harmonoise and CNOSSOS-EU outdoor sound prediction models are respectively known to have symmetric, asymmetric and omnidirectional sound emission directivities along front-back direction of the source vehicle. However, the influence of such difference in directivities to final predicted sound levels has not been investigated much. In this study, the influence is investigated using the ASJ Model and the Harmonoise under a configuration of semi-finite thick barrier along a source road. The configuration is an idealization of Japanese roadside buildings that have gaps in between, unlike European buildings that continuously extend over a whole urban block. Under the configuration, distribution of A-weighted sound levels around the end face of the barrier are computed with and without source directivity taken into account by each model. It is found from the results that the source directivity of the ASJ model makes little difference in the noise level distribution. In contrast, the source directivity of Harmonoise is found to make differences of 0.5-0.8 dB at right behind the barrier depending on vehicle running direction. However, a combined effect of source directivity and reflection at the end face is found to be negligible.